Google releases DIY open source Raspberry Pi 'Voice Kit' hardware -- here's how to get it

Two probably dodgy assumptions here:

- that you want an always-on microphone connected to the Googleplex in your living space. I don't.

- that a speech recognition system connected to a fixed range of applications merits being called an Artificial Intelligence.

As far as I'm concerned anything that is claimed to be an Artificial Intelligence has to be able to pass the Turing test as a bare minimum requirement. None of these things, whether from Apple, Amazon or Google even come close to doing that.

--
martin@   | Martin Gregorie 
gregorie. | Essex, UK 
org       |

I realise that any microphone connected to a computer is best treated as though it is "always on" but you do seem to have to press the button to talk to the AI-in-a-box.

Agreed.

The problem with that term is that it has *way* too many interpretations.

So the one that is currently the world's best Go player doesn't qualify ? How about the ones that generate pictures based on a short description ? Or the ones that learn languages and are currently being used to try and learn dolphin language ?

Passing a Turing test calls for either something designed for conversation or an artificial general intelligence. The latter is a current hot research target in AI, most of the older single purpose AI targets have been well and truly met now.

There's an odd thing about AI - every time some goal for AI development is reached there's a bunch of people explaining why the solution wasn't really AI just something else (Machine Learning is a popular name at the moment).

They don't even try, the main thing for them is reliable voice recognition regardless of accent or language with a decent tolerance for background noise. That's AI but the kind of single purpose AI that can be done very well these days.

I know a practical test for artificial self awareness - when an AI asks "what's in it for me" and means it that AI has to be treated as self aware.

--
Steve O'Hara-Smith                          |   Directable Mirror Arrays 
C:>WIN                                      | A better way to focus the sun 
The computer obeys and wins.                |    licences available see 
You lose and Bill collects.                 |    http://www.sohara.org/

Is that true of Google's Alexa equivalent when its on a phone or a PC? If not, are you sure the button isn't just added decoration on this thing?

Alexa, for one, doesn't do voice recognition and, as a result, has been triggered to buy something by a voice on the TV. At least twice.

Obligatory XKCD:

--
martin@   | Martin Gregorie 
gregorie. | Essex, UK 
org       |

It is on mine.

The Alexa in a Fire TV needs a button press to activate. (I don?t know about its other deployments, offhand.)

--
http://www.greenend.org.uk/rjk/

Definitely not. Just as chess programs and IBM's Watson do not - the latter has to be trained specifically for whatever you want it to do.

I'd agree that Watson and equivalent frameworks count as Expert Systems, but not as AIs.

Try this for size: to pass as an AI, a system must:

1) pass the Turing Test in more than one domain

2) when asked a question in a domain it has learned it must be able to (a)give a correct answer and (b) explain how it arrived at it

3) be able to learn how to play a game or to understand a technology by reading a book or manual (chess, go, C) or the game's instructions (monopoly, D&D)

4) be able to absorb more than one knowledge domain without getting confused.

Current systems can do 1 and 2a though I don't think anything can do both, and no current neural net-based system can do 2b. AFAIK nothing currently comes even close to tackling 3 or 4.

Can any make Celia Russell style word pictures? That should be quite easy since the word is part of the picture:

Are these just neural networks or something better and more flexible?

Quite and IMHO thats what the term should be reserved for.

I'd reverse that and say that whenever somebody has a new product that just a phrase->action mapper Siri, Alexa) or something a bit more interesting but that still can't explain how it reached the answer it gave, some idiot marketeer still calls it an AI.

--
martin@   | Martin Gregorie 
gregorie. | Essex, UK 
org       |

many usenet posters would fail that.

A majorette of usenet posters would fail that.

Few people can do that these days.

No one in politics would pass that.

The Rise of the Machines looms doesn't it?

--
The theory of Communism may be summed up in one sentence: Abolish all  
private property. 

Karl Marx

Ah these things are called general Artificial Intelligence these days. Hot research topic.

There is definitely progress being made on 3 and as a result 4, nothing by way of products but a good many papers and results.

Neural networks with memory and some fancy self training AFAICT.

Trouble is that leaves no term for things like the above which are not general AI but are also not programmed explicitly.

--
Steve O'Hara-Smith                          |   Directable Mirror Arrays 
C:>WIN                                      | A better way to focus the sun 
The computer obeys and wins.                |    licences available see 
You lose and Bill collects.                 |    http://www.sohara.org/

Once it masters that, we can move on to Scott Kim:

Anything to suck in the masses. Marketroids aren't quite the idiots that we might like to think they are.

--
/~\  cgibbs@kltpzyxm.invalid (Charlie Gibbs) 
\ /  I'm really at ac.dekanfrus if you read it the right way. 
 X   Top-posted messages will probably be ignored.  See RFC1855. 
/ \  HTML will DEFINITELY be ignored.  Join the ASCII ribbon campaign!

and cudent even spel "Touring Test".

That depends on which ng you're posting in.

Do you mean they can't use book learning or don't want to?

I hate to trying to learn anything from a video because the data rate is typically so much slower than reading text and because its difficult to stick bookmarks in a video.

In short, colour me totally bemused as to why anybody would choose to learn from a video when they can find a book or written instructions on the net. I suppose the only exception would be when its some really tricky manual operation and seeing it done beats looking at photos or diagrams.

Very true.

Still glowering distantly on the horizon: not nearly as looming as some would have us believe.

--
martin@   | Martin Gregorie 
gregorie. | Essex, UK 
org       |

That doesn't surprise me at all, though I don't like that term much as is sounds more than slightly premature.

Sounds good.

I though that would be the case.

I find this type of system a bit worrying because if you can't make a complex algorithm show you how it solved a problem it can be really hard to tell whether its answer is valid or to fix bugs.

Maybe better names would be Trainable Algorithm if it can't explain how it got its answer or Expert Algorithm if it can.

I think you have to use different names because whether an algorithm can explain itself should be a major factor when you're deciding how much you trust its output. After all, that's no different to how you decide whether to trust what a human tells you.

--
martin@   | Martin Gregorie 
gregorie. | Essex, UK 
org       |

AIUI the phone is listening locally for the "OK google" trigger phrase, then it starts listening properly and sending audio back to base to be decoded ... once or twice I've heard the phone say "If you just said something, I didn't catch what it was" when I haven't uttered the trigger phrase, as a coincidence it happened last night while I was listening to the radio, so I pressed rewind by a few seconds and thought I would tell what it had confused for the the trigger phrase, but it didn't trigger second time around.

Dunno, it might be able to be configured to listen 24x7, but the guy demoing it kept hitting the supplied button each time he spoke to it, guess it'll be in people's hands (from the magazine give-away) soon enough that people will be making unboxing videos about it ...

There is that - OTOH sometimes the answer is hard to find but easy to verify and in those cases something that can get there somehow will do just fine.

--
Steve O'Hara-Smith                          |   Directable Mirror Arrays 
C:>WIN                                      | A better way to focus the sun 
The computer obeys and wins.                |    licences available see 
You lose and Bill collects.                 |    http://www.sohara.org/

AFAICT the term has arisen to acknowledge that it's something that's not yet possible but being worked at and to distinguish it from the various flavours of single task AI which are getting pretty sophisticated these days.

This is true - although sometimes you trust a human based on experience of success rather than explanations, I expect the same will hold for general AI if/when it happens.

--
Steve O'Hara-Smith                          |   Directable Mirror Arrays 
C:>WIN                                      | A better way to focus the sun 
The computer obeys and wins.                |    licences available see 
You lose and Bill collects.                 |    http://www.sohara.org/

Cute, though I prefer M C Escher's stuff when it messing with perspectives, permutations, etc.

--
martin@   | Martin Gregorie 
gregorie. | Essex, UK 
org       |

"Six weaks ago I cuddnt even spel 'porgramr' - and now I are one."

alt.malaprop?

I'm with you. Text lets you proceed at my own pace, whether slower or faster than a video - and you can jump back and review earlier material at any time.

Agreed, there are cases where watching something helps everything fall into place in your mind.

I hear and I forget. I see and I remember. I do and I understand. -- Confucius

I'm less concerned about the machines than about the governments and megacorps that program them.

--
/~\  cgibbs@kltpzyxm.invalid (Charlie Gibbs) 
\ /  I'm really at ac.dekanfrus if you read it the right way. 
 X   Top-posted messages will probably be ignored.  See RFC1855. 
/ \  HTML will DEFINITELY be ignored.  Join the ASCII ribbon campaign!

Are their any guides to getting this working from a stock raspian start point? I want to try use my pi zero-w (posibly in a new case) without fully replacing its current setup.

--
COBOL is for morons. 
		-- E.W. Dijkstra

No guide but start from here

many thanks although I ended up using the google supplied Distro any way. All working without issue on a Pi Zero W :-)

--
FACILITY REJECTED 100044200000;